npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

@ktlsr/assay-adapters

v0.1.3

Published

Host adapters for Assay. The Claude Code adapter reads trigger, tool-trace, completion and cost signals from an isolated session.

Readme

@ktlsr/assay-adapters

Host adapters for Assay, a CI test runner for Agent Skills.

An adapter is the part that knows one host CLI: how to start a session, and how to read four signals back out of it — did the skill trigger, which tools were called, did the session actually complete, and what did it cost.

npm install @ktlsr/assay-adapters

Claude Code

ClaudeCodeAdapter is the adapter shipped today. It was chosen after a feasibility study across three hosts because all four signals are observable there without parsing prose: system/init reports the active skill set, a Skill tool_use announces a trigger, tool_result.is_error carries the trace signal, and the result message carries cost and latency.

import { ClaudeCodeAdapter } from '@ktlsr/assay-adapters'

const adapter = new ClaudeCodeAdapter()

Requires the claude CLI on PATH, plus CLAUDE_CODE_OAUTH_TOKEN (from claude setup-token) or ANTHROPIC_API_KEY.

Isolation

Every run gets its own temporary CLAUDE_CONFIG_DIR and the skill under test is loaded into that session alone. Without this you measure the machine's installation rather than the skill: in the experiment that produced this decision, an unisolated probe had 119 skills active, the target skill never fired on natural language, and the model reached for a neighbour's tool instead. Isolated, the same probe saw 19.

The child process does not inherit the parent environment. A short allowlist is passed through — PATH, home and temp, locale and timezone, proxy settings — plus the credential. Anything not on the list is absent, so a GITHUB_TOKEN in CI is not visible to the skill being measured.

The host is not taken at its word

finalize cross-checks the session before reporting completion. During the feasibility spike Claude Code twice reported subtype: "success" for runs that never happened — once "not logged in", once a revoked credential. The other fields told the truth. So a session with zero cost and zero output tokens, zero turns, or a terminal reason other than completed is marked unknown rather than trusted.

environmentHash derives a stable hash from the init fields — model, version, tool, skill, agent and plugin lists. It is reported as an environment hash, not a system prompt hash, because the host does not expose the system prompt and two different prompts could produce identical init fields. It is a real drift detector that claims less than it could.

License

Apache-2.0. See LICENSE and NOTICE.